Magnitude of Effects
Thus far, our vote count tallies suggest that only one community factor, disorder and incivilities, is more consistently associated with violent than nonviolent offending. It could still be the case that community factors are “significantly” associated with nonviolent offending, but the magnitude of effect is stronger for violent offending, and such a pattern would also support the differential etiology hypothesis. Several studies report standardized regression or correlation coefficients that enable us to contrast the strength of the effects of community factors on violent versus nonviolent crime in a limited way. The reader should be aware that a more formal test should be conducted before firm conclusions are drawn.
Across the structural factors that are predicted to raise levels of violence, there is no clear pattern suggesting that effects are stronger for violent than nonviolent crime (e.g., Bellair, 1997). Findings by Miethe et al. (1991) provide partial confirmation, in standardized coefficients suggest that ethnic heterogeneity is more strongly related to homicide and robbery rates than to the burglary rate in their cross-sectional analysis. (This was not true for change in crime rates.) Only Sampson and Groves (1989) find that the effects of ethnic heterogeneity on violent crimes are consistently stronger than the effects on nonviolent crimes, so the difference in magnitude does not support our hypothesis. Although we would not fully reject our hypothesis on this basis, there are too few studies to claim support for it.
The relative effect sizes for family disruption can be said to contradict the differential etiology hypothesis. In 2 studies, the percentage of female-headed households is more strongly related to violent than nonviolent crime (Lieberman & Smith, 1986; Steffensmeier & Haynie, 2000), and in 3 studies, the opposite was true (Hannon & DeFronzo, 1994; Oh, 2005; Sampson & Groves, 1989).
Earlier we reported that indicators of disorder were more consistently associated with violent than nonviolent crime. Just two studies reported the standardized coefficients needed to compare the magnitude of the effects. O’Shea (2006) reports mixed results in that an index of “social disorder” was more strongly related to the violent crime rate, while an index of “physical deterioration” was more strongly related to the property crime rate. On the other hand, Sampson and Raudenbush (1999) find that an index of observed social and physical disorder is more strongly related to violent crime than burglary.
In summary, based on a simple vote count analysis, the only evidence that community factors have a special relationship with violent crime comes from those studies which use indicators of physical or social disorder or incivilities. (The reader is reminded that community level poverty and concentrated disadvantage were covered in Chapter 9.) There are limitations to what we can glean from these studies. First, they vary in methodological quality, in particular with respect to statistical modeling. Second, modern social disorganization theorists posit that the effects of structural community factors on crime are indirect through their effects on neighboring and collective efficacy. There are too few studies testing this causal model for us to provide an evaluation here. Finally, quantitative research imposes severe constraints on the extent to which the reasons that community factors exert the effects that they do can be understood (e.g., Sampson, Morenoff, & Gannon-Rowley, 2002). Thus, in order to fully understand the complex, interdependent relationships between community factors and violence, we turn first to studies accounting for possible confounding factors, and then to the robust qualitative literature on neighborhoods and crime.